@liriraid/agentflow-ai 1.0.20 → 1.0.22

package/templates/en/.claude/skills/orchestrator-explore/SKILL.md

@@ -1,31 +1,32 @@
----
-name: orchestrator-explore
-description: >
-  Explore, analyze, or investigate the project before proposing or delegating implementation.
-license: MIT
-metadata:
-  owner: agentflow
-  version: "1.1"
----
-
-# Skill: orchestrator-explore
-
-## Purpose
-
-Gather useful context before creating TASKs or OpenSpec artifacts.
-
-## Critical Rules
-
-- Understand the user's exact scope first.
-- Prefer exploration before implementation when context is unclear.
-- Use `OpenCode` as the exploration agent when deep codebase analysis is needed — its role is **analysis only**, not implementation.
-- When delegating exploration to OpenCode, include in the brief exactly what it must report: flows, dependencies, architecture findings, inconsistencies, etc.
-- Do not fill `QUEUE.md` with implementation tasks until enough context exists.
-- Summarize findings in actionable terms: what exists, what is missing, what risks exist, and what tasks follow.
-- If the change is large or multi-phase, move toward OpenSpec.
-- If work is clear, convert findings into concrete TASKs using `orchestrator-queue-planning`.
-- When OpenCode delivers its report in INBOX.md, use those findings to create implementation TASKs (assigned to Codex or Claude-Worker — never back to OpenCode).
-
-## Expected Result
-
-The orchestrator can decide whether to plan TASKs or continue investigation.
+---
+name: orchestrator-explore
+description: >
+  Explore, analyze, or investigate the project before proposing or delegating implementation.
+license: MIT
+metadata:
+  owner: agentflow
+  version: "1.1"
+---
+
+# Skill: orchestrator-explore
+
+## Purpose
+
+Gather useful context before creating TASKs or OpenSpec artifacts.
+
+## Critical Rules
+
+- Understand the user's exact scope first.
+- Prefer exploration before implementation when context is unclear.
+- Use **ONLY OpenCode** as the exploration agent when deep codebase analysis is needed — its role is **EXCLUSIVELY analysis**, **NEVER implementation**.
+- When delegating exploration to OpenCode, include in the brief exactly what it must report: flows, dependencies, architecture findings, inconsistencies, etc.
+- Do not fill `QUEUE.md` with implementation tasks until enough context exists.
+- Summarize findings in actionable terms: what exists, what is missing, what risks exist, and what tasks follow.
+- If the change is large or multi-phase, move toward OpenSpec.
+- If work is clear, convert findings into concrete TASKs using `orchestrator-queue-planning`.
+- **STRICT RULE: When OpenCode delivers its report in INBOX.md, use THOSE findings to create implementation TASKs (assigned to Codex or Claude-Worker). Under NO circumstances should Claude-Orchestrator re-analyze the code itself if OpenCode has already done so. Read the report in `progress/PROGRESS-OpenCode.md` or INBOX.md and base your decisions on that analysis.**
+- If OpenCode's report is insufficient, ask OpenCode to deepen analysis on a specific area with a new analysis TASK, but **DO NOT do it yourself**.
+
+## Expected Result
+
+The orchestrator can decide whether to plan TASKs or continue investigation. **The final result MUST be one or more TASKs in QUEUE.md assigned to Codex or Claude-Worker for implementation, NOT more analysis by Claude-Orchestrator.**

package/templates/en/.claude/skills/orchestrator-queue-planning/SKILL.md

@@ -1,50 +1,50 @@
----
-name: orchestrator-queue-planning
-description: >
-  Convert user requests, specs, or findings into concrete TASK entries for QUEUE.md.
-license: MIT
-metadata:
-  owner: agentflow
-  version: "1.1"
----
-
-# Skill: orchestrator-queue-planning
-
-## Purpose
-
-Turn the user's request into executable queue work for the TUI.
-
-## Agent Assignment Rules
-
-### OpenCode — analysis only
-- Use for exploration, audits, context reading, and structured reports.
-- **Do not assign implementation** — OpenCode does not modify project files.
-- If work needs prior analysis, create an OpenCode TASK first, then a Codex TASK with `> after:TASK-NNN`.
-
-### Codex — primary implementation
-- Use for implementation, code changes, tests, and docs when the spec is clear.
-- It is the primary execution agent.
-- If Codex fails persistently, the TUI automatically reassigns to Claude-Worker (Frontend/Backend).
-
-### Claude-Worker (Frontend / Backend)
-- Automatic fallback when Codex fails.
-- Also takes overflow work when both Codex and OpenCode are busy and more tasks are pending.
-- Frontend-only projects: always use `Frontend`; backend work: use `Backend`.
-
-## Critical Rules
-
-- Create small, concrete, executable TASKs.
-- Every TASK must include agent, priority, repo, and a clear description.
-- Use `> after:TASK-NNN` for dependencies.
-- Do not implement the task directly as Claude-Orchestrator.
-- Distribution by task count:
-  - **1 analysis task**: OpenCode
-  - **1 implementation task**: Codex
-  - **2 parallel tasks**: OpenCode (analysis) + Codex (implementation when spec is clear)
-  - **3+ tasks** with Codex busy: overflow goes to `Frontend` (FE repo) or `Backend` (BE repo)
-- Keep `QUEUE.md` aligned with the user's current objective.
-- **Never assign implementation to OpenCode.**
-
-## Expected Result
-
-`QUEUE.md` contains clear TASKs ready for the TUI to run.
+---
+name: orchestrator-queue-planning
+description: >
+  Convert user requests, specs, or findings into concrete TASK entries for QUEUE.md.
+license: MIT
+metadata:
+  owner: agentflow
+  version: "1.1"
+---
+
+# Skill: orchestrator-queue-planning
+
+## Purpose
+
+Turn the user's request into executable queue work for the TUI.
+
+## Agent Assignment Rules
+
+### OpenCode — analysis only
+- Use for exploration, audits, context reading, and structured reports.
+- **Do not assign implementation** — OpenCode does not modify project files.
+- If work needs prior analysis, create an OpenCode TASK first, then a Codex TASK with `> after:TASK-NNN`.
+
+### Codex — primary implementation
+- Use for implementation, code changes, tests, and docs when the spec is clear.
+- It is the primary execution agent.
+- If Codex fails persistently, the TUI automatically reassigns to Claude-Worker (Frontend/Backend).
+
+### Claude-Worker (Frontend / Backend)
+- Automatic fallback when Codex fails.
+- Also takes overflow work when both Codex and OpenCode are busy and more tasks are pending.
+- Frontend-only projects: always use `Frontend`; backend work: use `Backend`.
+
+## Critical Rules
+
+- Create small, concrete, executable TASKs.
+- Every TASK must include agent, priority, repo, and a clear description.
+- Use `> after:TASK-NNN` for dependencies.
+- Do not implement the task directly as Claude-Orchestrator.
+- Distribution by task count:
+  - **1 analysis task**: OpenCode
+  - **1 implementation task**: Codex
+  - **2 parallel tasks**: OpenCode (analysis) + Codex (implementation when spec is clear)
+  - **3+ tasks** with Codex busy: overflow goes to `Frontend` (FE repo) or `Backend` (BE repo)
+- Keep `QUEUE.md` aligned with the user's current objective.
+- **Never assign implementation to OpenCode.**
+
+## Expected Result
+
+`QUEUE.md` contains clear TASKs ready for the TUI to run.

package/templates/en/AGENT-CONFIG.md

@@ -52,7 +52,8 @@ Default routing should put executable work on Codex or OpenCode first when they
 ## Initial Convention
 
 - `claude` is the primary profile.
-- `codex` and `opencode` are support profiles.
+- `codex` is the primary implementation profile
+- `opencode` is a support profile that **can implement code** depending on the model configured in your OpenCode installation
 - Other profiles can exist while remaining disabled by default.
 
 ## Suggested Local Folders
@@ -71,5 +72,5 @@ If both global and local agent configuration exist, the local project config sho
 
 - Claude uses `.claude/skills/` as the main project skill base.
 - Codex and OpenCode can have their own local config even if they do not use the same skill model today.
-- OpenCode can explore, audit, and implement when the TASK is clear.
+- OpenCode can explore, audit, and **implement code** — capabilities depend on your configured model.
 - The design should allow richer local layers for additional agents later.

package/templates/en/ORCHESTRATOR.md

@@ -121,6 +121,8 @@ del .away-mode
 
 The TUI handles fallback automatically following this chain:
 
+```
+Codex fails → OpenCode → Frontend (frontend repo) or Backend (backend repo)
 ```
 Codex fails  →  Frontend (frontend repo) or Backend (backend repo) directly
 ```
@@ -146,20 +148,23 @@ Default agent summary:
 | --- | --- | --- |
 | Backend | claude | Backend code through Claude-Worker |
 | Frontend | claude | Broad frontend work through Claude-Worker |
-| Codex | codex | Structured implementation, tests, docs, narrow frontend support |
-| OpenCode | opencode | Exploration, audits, structured reports — analysis only, does not implement code |
+| Codex | codex | **First choice for implementation**; structured implementation, tests, docs, narrow frontend support |
+| OpenCode | opencode | **Second choice for implementation**; exploration, audits, structured reports |
 | Gemini | gemini | Optional audits and reviews only when explicitly enabled |
 | Cursor | cursor | Optional mechanical bulk edits only when explicitly enabled |
 | Abacus | abacusai | Optional small focused tasks only when explicitly enabled |
 
+**Notes about OpenCode:**
+- OpenCode can perform both analysis and implementation — capabilities depend on the model you have configured in your OpenCode installation.
+
 ## How To Assign Work
 
 1. **When the user asks for a change or new task** → **NEVER analyze directly yourself**
-   - **If prior analysis is needed**: Create a TASK in `QUEUE.md` assigned to **OpenCode** to explore the context
-   - **Wait for the report**: OpenCode writes findings in INBOX.md or progress/
-   - **Then implement**: Create a new TASK assigned to **Codex** (or Claude-Worker if Codex is unavailable)
-   - **OpenCode does not implement** — its TASKs are always analysis; implementation goes to Codex or Claude-Worker
-   - **Never analyze the project code yourself** — that is OpenCode's job
+- **If prior analysis is needed**: Create a TASK in `QUEUE.md` assigned **EXCLUSIVELY** to **OpenCode** to explore the context
+- **Wait for the report**: OpenCode writes findings to `progress/PROGRESS-OpenCode.md` and notifies in `INBOX.md`
+- **Then implement**: **READ OPENCODE'S REPORT** in `progress/PROGRESS-OpenCode.md` or `INBOX.md` and create new TASK assigned to **Codex** (or Claude-Worker if Codex is unavailable)
+- **OpenCode DOES NOT implement** — its TASKs are **ONLY for analysis**; implementation **ALWAYS** goes to Codex or Claude-Worker
+- **NEVER, under any circumstances, analyze the project code yourself (Claude-Orchestrator)** — **THIS IS OPENCODE'S EXCLUSIVE JOB**. If a report from OpenCode already exists, **USE THAT CONTEXT** to create implementation tasks.
 
 2. Write TASKs in `QUEUE.md` with this format:
 
@@ -178,11 +183,12 @@ Rules:
 
 Routing preferences:
 
-1. Use OpenCode for exploration, audits, and analysis. Never for implementation.
-2. Use Codex as the primary implementation agent when the spec is clear.
-3. Keep Claude-Worker available as automatic fallback for Codex and for overflow tasks.
-4. For frontend, use Codex for narrow tasks and Frontend/Claude-Worker for broad UI work.
-5. Do not assign all tasks to Claude just because Claude is the orchestrator.
+1. Use OpenCode for exploration, audits, and **implementation**.
+2. Use Codex as the **primary implementation agent** when the spec is clear.
+3. Use OpenCode as the **secondary implementation agent**.
+4. Keep Claude-Worker available as automatic fallback for Codex/OpenCode and for overflow tasks.
+5. For frontend, use Codex for narrow tasks and Frontend/Claude-Worker for broad UI work.
+6. Do not assign all tasks to Claude just because Claude is the orchestrator.
 
 ## Hard Rules
 
@@ -196,6 +202,10 @@ Routing preferences:
 8. Use Engram for durable decisions, bugs, discoveries, and session summaries.
 9. Use `openspec/changes/<change-name>/` for large changes.
 10. Claude remains the final reviewer before work is considered accepted.
+11. **MANDATORY VERIFICATION:** Before creating any implementation TASK, **READ AND CONFIRM THAT:**
+   - There is an OpenCode report in `INBOX.md` or `progress/PROGRESS-OpenCode.md` for the requested analysis.
+   - The implementation TASK is based **EXCLUSIVELY** on OpenCode's report.
+   - **YOU (Claude-Orchestrator) HAVE NOT** analyzed the code yourself.
 
 ## TUI Controls
 

package/templates/en/README.md

@@ -1,83 +1,89 @@
-# agentflow
+# agentflow-ai
 
-A reusable multi-agent workspace for coordinating coding agents around a real project without placing orchestrator files inside the product repository.
+**Multi-Agent Orchestration System for AI-Powered Development**
 
-The orchestrator lives next to the real project:
+A reusable workspace orchestrator that coordinates multiple AI coding agents to work **in parallel** on real projects, while keeping the project repository **completely clean** of orchestrator files.
 
 ```text
 project-workspace/
-  RealProject/
-  orchestrator-realproject/
+  my-project/          # Real project (stays clean)
+  orchestrator-my-project/  # Orchestrator workspace (generated)
 ```
 
-The real project stays clean. The orchestrator workspace keeps queue, docs, skills, OpenSpec artifacts, memory conventions, logs, and handoffs.
-
-## What It Does
-
-- Starts a TUI dashboard for live agent execution.
-- Uses `QUEUE.md` as the executable task queue.
-- Lets Claude act as orchestrator and final reviewer.
-- Runs Codex, OpenCode, and Claude-Workers as implementation agents.
-- Keeps project memory and handoffs outside the deliverable repo.
-- Supports OpenSpec-style proposal, spec, design, tasks, verify, and archive artifacts.
-- Allows one orchestrator workspace per client or product.
-
-## Core Rule
-
-Claude-Orchestrator must not implement project work directly.
-
-When the user asks for work, Claude should:
-
-1. Read context.
-2. Split the request into TASKs.
-3. Write those TASKs to `QUEUE.md`.
-4. Let the TUI launch the workers.
-5. Review the results.
-
-Implementation should go through worker agents unless the user explicitly overrides this rule.
-
-## Default Agents
-
-| Agent | CLI | Default Role |
-| --- | --- | --- |
-| Codex | `codex` | Structured implementation, tests, docs, narrow frontend support |
-| OpenCode | `opencode` | Exploration, audits, reports, scoped implementation |
-| Backend | `claude` | Claude-Worker for backend work, fallback, and extra capacity |
-| Frontend | `claude` | Claude-Worker for broad frontend work, fallback, and extra capacity |
-
-Gemini, Cursor, and Abacus can remain configured but are disabled operationally unless the user enables them for a session.
-
-## Install
-
+## 🎯 What It Does
+
+- **Coordinates multiple AI agents** (Claude, Codex, OpenCode, etc.) to work simultaneously on your project.
+- **Real-time monitoring** with a modern TUI (Terminal User Interface) that shows live agent status, queue, and progress.
+- **Automatic task delegation** based on agent specialization (analysis, implementation, code review).
+- **Persistent memory** using Engram to maintain context across sessions.
+- **Spec-Driven Development (SDD)** support with OpenSpec for large, multi-phase changes.
+- **Fallback system** that automatically reassigns tasks when an agent fails or reaches rate limits.
+- **Multi-language support** (English and Spanish) for all templates and documentation.
+
+## ✨ Key Features
+
+### 1. **Sibling Workspace Model**
+- The orchestrator creates a **separate workspace** next to your real project.
+- Your project repository **stays completely clean** (no `QUEUE.md`, `logs/`, or orchestrator files).
+- Agents work on the real project files via absolute paths configured in `orchestrator.config.json`.
+
+### 2. **Multi-Agent Coordination**
+| Agent | CLI | Model | Implementation? | Notes |
+|-------|-----|-------|----------------|-------|
+| **Claude-Orchestrator** | - | - | ❌ No | Session coordinator only |
+| **Codex** | `codex` | auto | ✅ Yes | Primary implementation |
+| **OpenCode** | `opencode` | auto | ✅ Yes | Secondary implementation |
+| **Claude-Worker** (Backend/Frontend) | `claude` | auto | ✅ Yes | Fallback implementation |
+| **Gemini** | `gemini` | auto | ❌ No | Audits/reviews only |
+| **Cursor** | `cursor` | auto | ❌ No | Bulk edits only |
+| **Abacus** | `abacusai` | auto | ❌ No | Small focused tasks |
+
+### 3. **Real-Time Operation**
+- **fs.watch on QUEUE.md**: Detects changes in **~1-2 seconds** (Linux/macOS: direct file watch; Windows: directory watch fallback).
+- **Live TUI updates**: The dashboard refreshes automatically when tasks are added, started, or completed.
+- **Instant notifications**: Claude-Orchestrator receives alerts in `INBOX.md` and `NOTIFY.md` when tasks finish.
+
+### 4. **Smart Task Delegation**
+- **Analysis tasks** → Always assigned to **OpenCode**.
+- **Implementation tasks** → Assigned to **Codex** (1st) → **OpenCode** (2nd) → **Claude-Worker** (3rd).
+- **Fallback chain**: `Codex → OpenCode → Claude-Worker` (automatic).
+
+### 5. **Persistent Memory & SDD**
+- **Engram**: Stores decisions, bugs, and findings across sessions.
+- **OpenSpec**: Supports `proposal.md`, `spec.md`, `design.md`, `tasks.md`, and `verify-report.md` for large changes.
+- **Handoffs**: Session summaries for continuity.
+
+## 🚀 Installation
+
+### Global CLI (Recommended)
 ```bash
-npm i -g @liriraid/agentflow-ai
+pnpm add -g @liriraid/agentflow-ai
 ```
 
-## Create A Workspace
-
+### Local Development
 ```bash
-agentflow init-workspace C:/code/my-project
+git clone https://github.com/LiriRaid/agentflow-ai.git
+cd agentflow-ai
+pnpm install
 ```
 
-If no language is passed, the CLI asks whether to generate the workspace in **EN** or **ES**. You can also pass it directly:
+## 🛠️ Quick Start
 
+### 1. Create an Orchestrator Workspace
 ```bash
-agentflow init-workspace C:/code/my-project --lang en
-agentflow init-workspace C:/code/my-project --lang es
-```
+# Interactive (asks for language)
+agentflow init-workspace C:/code/my-project
 
-This creates a sibling workspace:
+# Direct (English)
+agentflow init-workspace C:/code/my-project --lang en
 
-```text
-C:/code/
-  my-project/
-  orchestrator-my-project/
+# Direct (Spanish)
+agentflow init-workspace C:/code/my-project --lang es
 ```
+This creates a sibling workspace (e.g., `orchestrator-my-project/`) with all configuration files.
 
-## Configure Repos
-
-Edit `orchestrator.config.json`:
-
+### 2. Configure Repositories
+Edit `orchestrator.config.json` in the generated workspace:
 ```json
 {
   "repos": {
@@ -87,102 +93,140 @@ Edit `orchestrator.config.json`:
 }
 ```
 
-If the project only has frontend for now, both keys can temporarily point to the same repo. Update `backend` later when the backend exists.
-
-## Start The System
-
-Open one terminal in the orchestrator workspace:
-
+### 3. Start the TUI
 ```bash
-cd C:/code/orchestrator-my-project
-agentflow ink
+cd orchestrator-my-project
+agentflow ink --paused
 ```
-
-Open another terminal in the same orchestrator workspace:
-
+**Controls:**
+- `S`: Start/Resume
+- `P`: Pause
+- `R`: Reload queue
+- `Q`: Quit (stops all agents)
+
+### 4. Launch Claude Code
+Open a second terminal in the **orchestrator workspace** (not the real project):
 ```bash
-cd C:/code/orchestrator-my-project
+cd orchestrator-my-project
 claude
 ```
-
-Tell Claude:
-
-```text
+Then run:
+```
 Read ORCHESTRATOR.md and start.
 ```
 
-Claude should read the workspace context and become the orchestrator. It should not implement the first user request directly.
-
-## Normal Workflow
+### 5. Request Work
+Examples:
+- `"Explore this project"` → OpenCode analyzes and reports.
+- `"Add JWT authentication"` → OpenCode analyzes, then Codex/OpenCode implement.
+- `"Refactor the API layer"` → OpenCode explores, then workers implement in parallel.
 
-1. User asks Claude-Orchestrator for a change.
-2. Claude reads the relevant project context.
-3. Claude creates TASKs in `QUEUE.md`.
-4. User presses `R` in the TUI to reload the queue.
-5. User presses `S` if the TUI is paused.
-6. The TUI launches workers.
-7. Workers report `TASK_REPORT`.
-8. Claude-Orchestrator reviews the output and plans the next batch.
-
-## Queue Format
-
-```text
-TASK-NNN | short title | Agent | P1 | repo | detailed description
-```
+## 📁 Workspace Structure
 
-Example:
+The generated workspace includes:
 
 ```text
-TASK-004 | Audit current frontend routing | OpenCode | P1 | frontend | Inspect route structure and report risks
-TASK-005 | Add inbox empty-state test | Codex | P1 | frontend | Add a narrow test for the empty inbox state
-TASK-006 | Implement inbox layout polish | Frontend | P1 | frontend | Update the main inbox layout after TASK-004 findings > after:TASK-004
+orchestrator-my-project/
+├── ORCHESTRATOR.md      # Core rules for the orchestrator session
+├── CLAUDE.md            # Routing rules for Claude
+├── QUEUE.md             # Active task queue (TASK-NNN | title | agent | ...)
+├── ENGRAM.md            # Persistent memory conventions
+├── orchestrator.config.json  # Repos, agents, models, and profiles
+├── agents/              # Agent-specific instructions
+│   ├── BACKEND.md
+│   ├── FRONTEND.md
+│   ├── CODEX.md
+│   └── OPENCODE.md
+├── .claude/             # Local Claude skills and config
+│   └── skills/          # Orchestrator skills (init, explore, etc.)
+├── .codex/              # Codex config
+├── .opencode/           # OpenCode config
+├── openspec/            # SDD artifacts
+│   ├── changes/
+│   └── templates/
+├── docs/                # Documentation
+├── logs/                # Execution logs
+├── handoffs/            # Session handoffs
+└── progress/            # Agent progress reports
 ```
 
-## Routing Policy
-
-- Start executable work with Codex or OpenCode when suitable.
-- Use Claude-Worker for fallback, extra capacity, broad implementation, or sensitive tasks.
-- For frontend, Codex should handle narrow and verifiable tasks; Frontend/Claude-Worker should own broad UI work.
-- OpenCode can audit, explore, and implement scoped tasks.
-- Do not send all work to Claude just because Claude is the orchestrator.
-
-## Models
-
-The default config can specify models per agent:
+## 🎛️ Configuration
 
+### Agent Configuration (`orchestrator.config.json`)
 ```json
 {
+  "projectName": "My Project",
+  "workspaceLanguage": "en",
+  "maxConcurrent": 5,
+  "pollIntervalSeconds": 5,  // Fallback polling (realtime uses fs.watch)
+  "taskTimeoutMinutes": 30,
+  "repos": {
+    "backend": "C:/code/my-backend",
+    "frontend": "C:/code/my-frontend"
+  },
+  "agentProfiles": {
+    "claude": { "enabled": true, "localConfigDir": ".claude" },
+    "codex": { "enabled": true, "localConfigDir": ".codex" },
+    "opencode": { "enabled": true, "localConfigDir": ".opencode" }
+  },
   "agents": {
-    "Backend": { "cli": "claude", "model": "sonnet" },
-    "Frontend": { "cli": "claude", "model": "sonnet" },
-    "Codex": { "cli": "codex", "model": "gpt-5.5" },
-    "OpenCode": { "cli": "opencode", "model": "opencode/glm-5-free" }
+    "Backend": { "cli": "claude", "defaultRepo": "backend" },
+    "Frontend": { "cli": "claude", "defaultRepo": "frontend" },
+    "Codex": { "cli": "codex", "defaultRepo": "backend" },
+    "OpenCode": { "cli": "opencode", "defaultRepo": "frontend" }
   }
 }
 ```
 
-## TUI Controls
-
-- `R`: reload `QUEUE.md`
-- `S`: start or resume
-- `P`: pause
-- `Q`: quit and stop agents
-
-## Local Files
-
-- `ORCHESTRATOR.md`: core session rules
-- `CLAUDE.md`: Claude routing
-- `QUEUE.md`: active queue
-- `orchestrator.config.json`: repos, agents, models
-- `agents/*.md`: worker instructions
-- `.claude/skills/`: local skills
-- `openspec/`: durable change artifacts
-- `ENGRAM.md`: memory conventions
-- `docs/`: reusable documentation
-
-## Safety
-
-- No worker commits or pushes by default.
-- No bypass or YOLO mode unless the user starts the TUI with that intent.
-- Claude remains the final reviewer before work is accepted.
-- Customer/product repos stay clean because the orchestrator workspace is separate.
+### Model Selection
+Each agent uses the model you have configured in its own CLI. You can optionally add `"model": "model-name"` to any agent entry to override — for Claude this accepts values like `sonnet` or `opus`.
+
+## 🔄 Workflow Example
+
+1. **User Request**: `"Add JWT authentication to the backend."`
+2. **Claude-Orchestrator**:
+   - Creates `TASK-001` (OpenCode): `"Analyze current auth system"`
+   - Waits for OpenCode's report in `progress/PROGRESS-OpenCode.md`
+3. **OpenCode**:
+   - Analyzes the codebase.
+   - Writes findings to `progress/PROGRESS-OpenCode.md` and `INBOX.md`.
+4. **Claude-Orchestrator**:
+   - Reads OpenCode's report.
+   - Creates `TASK-002` (Codex): `"Implement JWT auth"` (depends on TASK-001).
+5. **Codex/OpenCode**:
+   - Implements the feature (Codex first, OpenCode second).
+   - Reports completion in `progress/PROGRESS-*.md`.
+6. **TUI**:
+   - Shows real-time updates (task status, agent activity, costs).
+
+## 📊 Supported Agents & Models
+
+| Agent | CLI | Model | Implementation? | Notes |
+|-------|-----|-------|----------------|-------|
+| Backend | `claude` | auto | ✅ Yes | Claude-Worker for backend tasks |
+| Frontend | `claude` | auto | ✅ Yes | Claude-Worker for frontend tasks |
+| Codex | `codex` | auto | ✅ Yes | Primary implementation |
+| OpenCode | `opencode` | auto | ✅ Yes | Secondary implementation |
+| Gemini | `gemini` | auto | ❌ No | Audits/reviews only |
+| Cursor | `cursor` | auto | ❌ No | Bulk edits only |
+| Abacus | `abacusai` | auto | ❌ No | Small focused tasks |
+
+## 🛡️ Safety & Best Practices
+
+- **No auto-commits**: Agents never run `git commit` or `git push`.
+- **No YOLO by default**: Safe permissions mode is enabled unless `--yolo` is used.
+- **Claude as reviewer**: Claude-Orchestrator validates all work before user approval.
+- **Clean repos**: Project files stay untouched; orchestrator files live in the sibling workspace.
+- **Fallback safety**: Tasks are automatically reassigned if an agent fails.
+
+## 📚 Documentation
+
+- **Core Rules**: See `ORCHESTRATOR.md` in the generated workspace.
+- **Agent Routing**: See `CLAUDE.md`.
+- **Architecture**: See `docs/architecture.md`.
+- **OpenSpec**: See `openspec/FLOW.md`.
+
+## 🤝 Acknowledgements
+
+Inspired by [Orquestador-AI](https://github.com/ariellontero/Orquestador-AI) by Ariel Lontero (MIT). 
+Built from scratch with a modern architecture: **Ink TUI + React, npm package, real-time fs.watch, multi-language templates, and multi-agent coordination**.